HybridLT算法的改进及其收敛性分析

Improvement and convergence on the Hybrid LT scheme

HybridLT法是一种解决组合优化问题的新方法 .与现存的Hopfield网络比较 ,这种方法有几个优点 :它不仅能用于二次函数 ,还能用于非二次函数 ;另外 ,在控制人为的加权参数时 ,它减少了对外部的依赖 .文中对原算法进行了改进 ,使其能更快找到准确的最优点 ,并且对迭代方法做了收敛性分析 ,给出了收敛的必要条件和充分条件 .从而说明HybridLT方法是可行的 ,是一种适用于很多类型组合优化问题的有效方法 .

The Hybrid LT scheme is a new type of scheme for combinatorial optimization problems. Comparied with the existing Hopfield type networks, the new scheme has several new features. It is applicable to nonquadratic energy functions, and externally artificially controlling weighting parameters are easily controlled. In this paper, the Hybrid LT scheme is improved. The convergence on the improved Hybrid LT scheme is investigated, and the necessary and sufficient conditions for convergence are proposed. By the analysis of these convergent conditions, the improved Hybrid LT scheme is revealed to be efficient and convergent for a very wide variety of combinatorial optimization problems.